Don’t know if anyone said this before, but there is the issue of font at play too. Your brain, if you only use the srs, isn’t learning to process the abstraction of the kanji, aka, the generalization of the shape of the kanji.
In other words, even though there are slight differences between the ways the kanji looks in different fonts, it is still the same kanji. So when you are looking at the kanji in a new font, one different from the bunpro font, your brain says to itself, “I know what the ‘読’ kanji looks like, and what I’m looking at is slightly different, so I know that it isn’t ‘読’. Wow! I have no idea what this kanji is, I’ve never seen it before, but I at least know it’s not ‘読’.”
That being the case, your subconscious brain can’t tell you what it is you are looking at because it is sure it has no idea what this ‘読’ like thing it is looking at is. You might think that I’m being silly, but the brain is incredible powerful at comparison and recognition, and it is literally doing what I’m saying it is doing.
So until your brain sees 読 in several different fonts, and learns an abstraction of 読 that it can apply in several different situations, you won’t be able to recall 読 is when you see it in the wild, because your brain has no idea what it is processing.
Bunpro should really include and use several different fonts automatically on the site so you learn these abstractions right away. Just my two cents.